Framing Processes in Social Dilemmas. Formal Modelling and Experimental Validation. Christian...

Framing Processes in Social Dilemmas.

Formal Modelling andExperimental Validation.

Christian Steglich, ICS Groningen

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Purpose of the presentation

is the validation of (some of) LINDENBERG’s ideas about the microfoundations of solidarity

as they are e.g. spelled out in Chapter 3 of

Doreian & Fararo (eds.):

The Problem of Solidarity: Theories and Models,

Amsterdam 1998 (Gordon & Breach).

How?• The theory is an application of framing theory.• Thus: test it by means of “framing analysis.”• Take social dilemmas as test domain.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

A group faces a social dilemma when the following two properties hold (DAWES 1980):

each group member is worse off when (s)he cooperates

than when (s)he defects, irrespective of what the other group members do:

each group members is better off when everyone

cooperates than when everyone defects:

: ( | : ) ( | : )i i j i i ji v c j c v d j d

: ( | .) ( | .)i i i ii v c v d

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

A group faces a social dilemma when the following two properties hold (DAWES 1980):

each group member is worse off when (s)he cooperates

than when (s)he defects, irrespective of what the other group members do:

each group members is better off when everyone

cooperates than when everyone defects:

: ( | : ) ( | : )i i j i i ji v c j c v d j d

: ( | .) ( | .)i i i ii v c v d

dilemma

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules,

• introduce punishments for defection,

• appeal to farsightedness.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules,

• introduce punishments for defection,

• appeal to farsightedness:

GAME THEORY.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules,

• introduce punishments for defection,

• appeal to farsightedness:

GAME THEORY.

not here

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules,

• introduce punishments for defection:

SANCTIONING SYSTEMS.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules,

• introduce punishments for defection:

SANCTIONING SYSTEMS.How are these provided?

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules:

NORMS.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

How to solve the social dilemma?

Suggestions in the literature:

• coordinate behaviour by obligatory rules:

NORMS.How does obligation work?

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Normative behaviour …

… is always stabilized by sanctions.

Absence of sanctions is a telltale sign that a behavioural rule is not normative.

… is internalized .Actors want to do what they have to do.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Sanctions …

… are integral part of normative behaviour.

Absence of sanctions is a telltale sign that a behavioural rule is not normative.

… but do not (directly) influence behaviour.

Actors want to do what they have to do.

(the “sociologists’ dilemma” )

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

LINDENBERG’s theory of norms:

• when sanctions directly influence

behaviour, the actor “is in a gain frame.”

(foreground influence of sanctions)

• when an actor “is in a normative frame,”

sanctions only influence the strength of the norm, not its content.

(background influence of sanctions)

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

LINDENBERG’s framing theory(Discrimination model of

framing):

• weakness of a frame leads to random

preference, and vice versa. • weakness of a frame leads to a frame switch.

Experimental validation will centre around

the dynamic manipulation of frame strength by variation of sanction sizes.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Some hypotheses:

• sensitivity to sanction size differs between frames:

normative frame: lower sensitivity,

gain frame: higher sensitivity.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Some hypotheses:

• sensitivity to sanctions differs between frames,

• attitude towards sanctions differs between frames:

normative frame: positive attitude,

gain frame: negative attitude.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Some hypotheses:

• sensitivity to sanctions differs between frames,

• attitude towards sanctions differs between frames,

• behavioural randomness depends on frame×sanction

interaction:

normative frame: behavioural randomness

occurs for low sanctions,

gain frame: behavioural randomness

occurs for high sanctions.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Some hypotheses:

• sensitivity to sanctions differs between frames,

• attitude towards sanctions differs between frames,

• behavioural randomness depends on frame×sanction

interaction,

• stability of frames over time:Actors approach decision

situations with theframe they applied in the

previous situation.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Some hypotheses:

• sensitivity to sanctions differs between frames,

• attitude towards sanctions differs between frames,

• behavioural randomness depends on frame×sanction

interaction,

• stability of frames over time:- inertia of frames and

behaviour,- hysteresis of frames and

behaviour.

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Formal modelling:

• Assume that before making a decision in a social

dilemma, actors adopt either a normative or a gain frame: F {fnorm, fgain} .

This framing stage is influenced by situational

parameters s and the previously used frame.

• Assume that then, actors base their behaviour on

a frame-dependent decision rule:Y ~ (s|F) .

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Formal modelling:

The model can be summed up visually as follows:

Fn-1

sn-1

Yn-1

Fn

sn

Yn

Fn+1

sn+1

Yn+1

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

The experimental study:

(January 2001, 124 students, computer experiment.)

Task: Protection of wild animals over N=21 days,

• cooperation was tied to a reduction in

collective housing costs,

• defection meant private gain (and was sanctioned by percentage s).

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Experimental conditions:

• sanctioning pattern: \/ versus /\ ,

• semantic framing a (for accessibility manipulation):

environmentalist group versus leisure time brokers.

Dependent variables:

• sanctioning attitude x (adequate sanctions in %),

• contribution y to common task (in hours out of 10h).

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Analytical framework:

• initial frame probabilities:

• rules for frame updating:

• rules for frame-dependent behaviour:Y ~ beta(p,q) with p mean contribution:

and q corrected variance:

0 0 1[ Pr( )] normlogit F f a

1 0 1 2 3 1 4[Pr( | )] n n nm m m m m m mlogit F f F f a s y n

1 2( ) of f f flogit p a s

1 2( ) of f f flogit q a s

Descriptive results: gain semanticsnormative semantics

patt

ern

\/pa

tter

n /\

Descriptive results: hysteresis hypothesis confirmed. gain semanticsnormative semantics

patt

ern

\/pa

tter

n /\

hysteresis

Descriptive results: semantic framing successful. gain semanticsnormative semantics

patt

ern

\/pa

tter

n /\

semantics

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Model estimates:

Behavioural rules per frame

1007550250

1.0

.8

.6

.4

.2

0.01007550250

1.0

.8

.6

.4

.2

0.0

sensitivity hypothesis confirmed

behavioural variation hypothesis confirmed

gain frame

normative frame

sanction size (in %)

mea

n co

ntri

buti

on p

corr

ecte

d va

rian

ce q

Model estimates: Distribution of contributions per frame

norm

ativ

e fr

ame

gain frame

no sanctions

50 % sanctions

100 % sanctions

Model estimates: once more behavioural variation

norm

ativ

e fr

ame

gain frame

no sanctions

50 % sanctions

100 % sanctions

non-discriminati

on

non-discrimination

Model estimates: “rationality” of gain frame’s rule ?

norm

ativ

e fr

ame

gain frame

no sanctions

50 % sanctions

100 % sanctions

¿ model artefact or background effect ?

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Model estimates:

Frame updating: regions of frame stability

gain frame

normative frame

The inertia hypothesis is partly confirmed by threshold shape : Frames are stable in the region of compatible behaviour.

previous contribution in hours

1050

fram

e re

plic

atio

n pr

obab

ility

1.0

.8

.6

.4

.2

0.0

previous contribution in hours

1050

fram

e re

plic

atio

n pr

obab

ility

1.0

.8

.6

.4

.2

0.0

normative sem

anticsgain sem

antics

previous contribution in hours

fram

e re

plic

atio

n pr

obab

ility

Model-derived simulations: goodness of fit visualised. gain semanticsnormative semantics

patt

ern

\/pa

tter

n /\

Model-derived simulations: fit problemsgain semanticsnormative semantics

patt

ern

\/pa

tter

n /\

Simulated contributions less extreme than in data.

Model-derived simulations: fit problemsgain semanticsnormative semantics

patt

ern

\/pa

tter

n /\

Contributions increase for small, decreasing sanctions (gain frame).

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

External validity of the model:

The estimates are solely based on actors’ behaviour Y.Model-derived frames can now be compared to the other dependent variable sanction attitude X:

The sanction attitude hypothesis is confirmed.

sanction attitudepositive negative

normative 1129 183gain 198 1094es

timat

edfr

ame

presentation prepared for the Mathematical Sociology meeting, 27 November 2002

Conclusions:

• Framing theory gives a valid account of behaviour.

• The theory of normative behaviour is confirmed.

• The model-fitting procedure does a good job

(but suffers from rigid specifications).